The Dilemma of Halting Clinical Trials Early
http://www.medpagetoday.com/HematologyOncology/ClinicalTrials/tb/2047

By Neil Osterweil, Senior Associate Editor, MedPage Today
Reviewed by
Zalman S. Agus, MD; Emeritus Professor at the University of Pennsylvania School of Medicine.
November 01, 2005

MedPage Today Action Points

• Inform interested patients that there could be a variety of reasons -- some clinical, some ethical, some financial -- why a randomized clinical trial may be stopped early for benefit.
  
  
• Inform patients that in well-designed trials a data-safety monitoring board carefully analyzes all relevant data to ensure patient safety.

Review
HAMILTON, Ontario, Nov. 1 - The results of clinical trials that are halted early because of a reported dramatic benefit for the treatment group should be viewed with a jaundiced eye, suggest researchers here.

At first glance, it might seem unethical to continue with a randomized trial that involves withholding from one group a treatment that has clearly benefited others.

Yet, truncating a trial robs researchers of the chance to learn all they can about the longer-term efficacy of a given therapy, and it may put patients at risk for unexpected adverse events that crop up over time, point out Gordon H. Guyatt, M.D., M.Sc., a professor of clinical epidemiology and biostatistics at McMaster University and colleagues here.

In addition, because the statistical base is smaller when a trial is cut short for benefit, the magnitude of the beneficial effect may be exaggerated. And it surely doesn't hurt marketing efforts when drug makers can claim that because their products proved to be so good, it was their solemn duty to offer them as rapidly as possible.

On the other hand, say investigators involved in early-halted trials, there are clearly cases where it's downright irresponsible not to offer a given treatment, such as a life-saving cancer therapy or heart drug, to all eligible patients. A mound of data never equates with a human life, they argue.

"Randomized clinical trials stopped early for benefit are becoming more common, often fail to adequately report relevant information about the decision to stop early, and show implausibly large treatment effects, particularly when the number of events is small," wrote Dr. Guyatt and colleagues here in a review article in the Nov. 2 issue of the Journal of the American Medical Association.

"These findings suggest clinicians should view the results of such trials with skepticism," they added.

"If I had my way, we would never stop early," Dr. Guyatt said in an interview. "I think in general it is a mistake, That is an extreme position, but it's actually what I believe would be best for the clinical trials endeavor and for the patients."

But a leading investigator of the ASCOT-BPLA trial, which was halted and unblinded early because of significantly higher all-cause mortality in one arm of the trial, says that while continuing that study might have satisfied the statisticians, it would have been bad for patients and for investigators.

"Would you go to court and say you carried this on for another two years knowing that the death rate was higher in one side than in the other?" asked Neil Poulter, M.D., a professor of preventive cardiovascular medicine at Imperial College in London, in an interview.

"This isn't some wussy endpoint that you can say is counterbalanced by something else -- this is death," Prof. Poulter continued. 'Who would say 'Yep, sure, fine carry on for another two years'?"

ASCOT-BPLA, the Anglo-Scandinavian Cardiac Outcomes Trial-Blood Pressure Lowering Arm (ASCOT-BPLA) showed that the combination of Norvasc (amlodipine) and Aceon (perindopril) is an effective therapy, compared with the older beta-blocker atenolol with or without bendroflumethiazide.

The trial was halted early in December 2004 when the data safety monitoring board determined that there was a higher event rate in the atenolol arm, including a 24% greater incidence in all-cause morality.

The decision to stop the trial created a bias that exaggerated positive findings, said Salim Yusuf, M.D., director of cardiology at McMaster University here, when the study was discussed at the European Society of Cardiology Meeting in September.

It isn't always so cut and dried, however. In fact, says a member of the data safety monitoring board for the ASCOT-BPLA study, the DSMB originally recommended the possibility of ending the trial early in October of 2003.

"We did make a recommendation about stopping at that point, but we do not have decision powers, as is good practice -- we just make a recommendation," said Stuart Pocock, Ph.D., a professor of medical statistics and director of clinical trials research at the London School of Hygiene and Tropical Medicine at the University of London. Dr. Pocock wrote an editorial accompanying today's JAMA review.

Because the differences in ASCOT-BPLA seen at an early interim analysis could have been chalked up to differences in blood pressure between the trial groups that might have evened out over time, the investigators decided to continue the trial, Prof. Poulter said.

But one year later when the monitoring board members saw the all-cause mortality data, they called for an immediate halt to the study.

"Now, five people -- five totally independent, highly-rated people, some of the top people in the world -- said to us 'You have to stop this trial, this is not right,'" Dr. Poulter said. "So we then put it to the steering committee, which has got one person from every country, and every single one was unanimous in saying if we didn't stop the trial, we couldn't look at ourselves in the mirror."

It's a compelling argument. But when the gates of secrecy protecting the data in randomized clinical trials are prematurely flung open, and the treatment alleged to be superior is offered to all patients in the study, the chance for long-term, objective evaluation of the drug or therapy in question may be lost forever, Dr. Guyatt and colleagues contended.

In addition, trials that are stopped early, when relatively few clinical endpoints have occurred, may exaggerate the ultimate benefit of a therapy, they determined.

Dr. Guyatt and colleagues found that in a review of 143 randomized clinicals trials "on average, the trials recruited 63% (SD, 25%) of the planned sample and stopped after a median of 13 (interquartile range [IQR], 3-25) months of follow-up, one interim analysis, and when a median of 66 (IQR, 23-195) patients had experienced the end point driving study termination."

In his editorial, Dr. Pocock noted that in the CHARM trial, comparing the drug Atacand (candesartan) with placebo for heart failure, there was a temptation to stop at the fourth interim analysis, when a 24% reduction in the risk of death was seen for the candesartan group. The difference at the time crossed the pre-determined stopping boundary of P<.001. The monitoring board for that trial decided to continue to the next interim analysis, and the next, until the original goal of 3.1 median years of follow-up was reached, at which time they found that the benefit of Atacand had shrunk to 9% (P=.055).

"Early stopping was resisted, and hence any exaggerated claim of survival benefit was avoided, and important long-term benefits in other outcomes, such as cardiovascular death and heart failure hospitalization, were realized in each of the three component trials of CHARM," Dr. Pocock wrote.

In an interview, Dr. Guyatt acknowledged that "there is legitimate concern about protecting the people who are involved in the trial, and so not subjecting people to the option of a placebo when there is a dramatically beneficial treatment."

"However," he added, "it seems to me that investigators and data safety monitoring boards have lost sight to some extent of their responsibilities to people outside of the trial. So if a trial is highly publicized with a misleading estimate of benefit, there are going to be thousands, tens of thousands, hundreds of thousands, perhaps even millions of people who are presented with misleading information that may lead them to make poor choices, and so I think that the ethical obligations to those individuals should be something that investigators and data safety monitoring boards take very seriously."

In the case of the ASCOT-BPLA trial, however, the trial was stopped only after more than 500 events had occurred, Dr. Poulter noted.

Another issue that concerns critics is that when a big clinical trial is halted early for benefit, there's often a whiff of showmanship and the appearance of marketing in the way the results are announced. Results of 92 of the 143 trials Dr. Guyatt and colleagues reviewed cropped up in five "high-impact" medical journals: New England Journal of Medicine, JAMA, The Lancet, Annals of Internal Medicine and the British Medical Journal.

What's more, the authors found, a large proportion of the trials are funded by the drugs and devices industry, and involve drugs with large market potential in fields such as cardiology, oncology, and HIV/AIDS.

"There are a whole bunch of reasons that various parties in the exercise may like to stop early," Dr. Guyatt said. "If effect sizes are inflated, at least in terms of profit it's actually in the interest of the industry. Stopping early means that you devote fewer resources to the trial that you can devote elsewhere."

"Stopped early results with big effects tend to be published, as we demonstrated, in the very top journals, grossly disproportionately, and it gives ready-made publicity," he added "From the investigators point of view it gets them publication in the top journal and potential invitations to address audiences in places that may be attractive to them."

Dr. Guyatt and colleagues also take issue with the fact that many of the trials they included in their review did not report the planned sample size, the interim analysis after which the trial was stopped, or whether a stopping rule or "adjusted analysis accounting for interim monitoring and truncation," they noted.

Ultimately, the question of whether to stop or solider on comes down to a question of ethics.

"On the one hand, if early on in the trial you see evidence of a treatment difference, you then have this ethical concern that future patients who enter that trial may be allocated to what is in truth an inferior treatment," Dr. Pocock said.

"On the other side of the dilemma, you want to make you get the right answer in the long run, with lots of evidence, so that you really change future practice for the better," he added.

Primary source: Journal of the American Medical Association
Source reference:
Montori VM et al. Randomized Trials Stopped Early for Benefit. JAMA. 2005;294:2203-2209

Additional source: Journal of the American Medical Association
Source reference:
Pocock SJ..When (Not) to Stop a Clinical Trial for Benefit. JAMA. 2005;294:2228-2230.